Fighting The Demon

Is a virus a strategic actor? Most would say no. Many find the entire notion to be totally absurd. In wars the enemy is an active opponent capable of reasoning about goals and pursuing them. A virus is far more governed by impersonal mathematical laws than anything else. Inasmuch as it can improve and adapt its behaviors, it does so via evolutionary processes that select for behaviors that grant it increased biological fitness. All of this is of course less about viruses and biological arcana and more about how we understand ourselves. Humans are not free of hard biological constraints. However, we are given a “long leash” to flexibly pursue goals in a changing environment. We are aware of our chains, and sometimes we can break free of them.

Human beings, we tell ourselves, are distinct due to our capacity for mental rather than purely motivational autonomy. If motivational autonomy is the ability to flexibly select actions that fulfill particular needs, mental autonomy is the ability to reflect on and change those underlying needs. This activity is informed by both internal human mental capacities as well as interaction with external forms of thought and memory that can aid individual and collective deliberation. Hence we give ourselves a rather exalted place in the strategic hierarchy. But one of the long-run impacts of biological science has been to collapse the old tradition of Great Chains of Being that place humans at the top of the pecking order. And with the novel coronavirus the psychological effects of this dethroning once again are quite severe.

Venkatesh Rao describes the problem coronavirus poses to us as one of shock and horror that an inhuman entity has put us all on the defensive.

As far as we can tell, a virus has gotten inside the [Observe, Orient, Decide, Act] loop of Homo sapiens, and seized the initiative, while we’re struggling to figure out what to even call it. It is spreading faster than our fastest truths, lies, and bullshit. A supersonic shock wave in the narrative marketplace. There is a sudden recalibration of how much agency humanity collectively possesses, and we’re not happy with the results. … it feels a little silly to declare “war” on COVID-19 or declare that we will “defeat” it. It doesn’t know it is playing or competing. There is nothing it is like to be COVID-19, let alone COVID-19 in a winning/losing mental state. At best it can be cast as the instrument of an angry god. For us atheists, it’s a molecule that’s barely qualifies as living. … In OODA loop theory… [you] get inside [the adversary’s] decision cycle and collapse it from within, so they can no longer tell significant/insignificant apart, and their psyche collapses under the stress of maintaining orientation and continuing to act meaningfully. The hardest part of OODA theory is accepting that this can be done to us by adversaries that are not necessarily smarter or more capable than us; they need only lack our illusions.

Part of the problem may be differences in how we think about strategic behavior across disciplines. The strategic thinking of biologists is very different from strategy as traditionally understood in other fields. Growing out of complicated interdisciplinary Cold War collaborations, biologists eventually stripped conscious rational agency out of strategic behavior. Richard Dawkins, Paul W. Ewald, and other biologists interpret entities like pathogens in strategic terms. Tumor cells “cooperate” to kill you and so forth. Due to the power of “economic imperialism”, modelling formalisms that originated to describe human behavior migrated to other fields – albeit mutating in the process. Hence the unusually actor-centric language that biologists sometimes use.

To those that lack at least casual familiarity with scientific developments in the biological sciences since the mid 20th century, this language can be somewhat strange to assimilate. Nonetheless, it is part of a longstanding tradition that merges together game theory and biology to interpret strategies as an organism’s genetically determined characteristics and behaviors. If strategy is the organism’s phenotype and fitness is the payoff, strategic phenomena is primarily about the frequency of strategies within a population of players in the game. Unsurprisingly there is a very long history of using computers to simulate strategic behaviors using these stylized assumptions. Individual organisms can be programs with hard-coded behaviors that only change over time within the population rather than their own individual lifetimes.

The speed of pathogenic evolution mitigates the inflexibility of individual pathogenic behaviors. In his book on the evolution of infectious diseases (recommended by Dawkins somewhat cryptically with the tagline “think of virus as strategist”), Ewald goes into detail about the complicated tradeoffs that shape pathogenic evolution. Some of this is chilling. Ewald dismantles a common assumption – at least when the book was released in the 1990s – that it is in the interest of parasites to evolve towards some kind of peaceful coexistence with parasite hosts. More optimistically, Ewald also suggests that it is possible to potentially shape the evolution of pathogens towards less harmful variants using public health interventions. But most importantly, Ewald repeatedly emphasizes that evolutionary processes that ordinarily occur on far longer time scales are blindingly fast in pathogenic evolution.

The evolution of human characteristics is relatively slow… Pathogens, in contrast, may evolve substantially over time periods of a few weeks. When considering infectious diseases from the pathogen’s “point of view,” a few decades of medical records may offer a potential for evolutionary change that is comparable to the entire time span of our genus Homo Sapiens. The evolutionary process for pathogens is therefore best considered to be a process in progress. The pathogens are a moving target of our research. We and our activities are part of the environment that pushes this process down one course or another… My emphasis will be from the parasite’s point of view rather than the hosts. It is not that I have any fondness for [the parasites]. No, I emphasize the parasites’ point of view because the health sciences have strongly emphasized the humans’ point of view.

Of course, there is always risk in using the language of rational and agential behavior to describe entities that philosophers and scientists argue may or may not be alive in the first place. There is a significant difference between whether or not it makes sense for an external observer to ascribe characteristics to a particular thing being observed and whether or not that thing can be said to have those characteristics and meaningfully deploy them. This is why we perhaps ought to go back to older – and more sophisticated – language that bypasses many of these concerns. Perhaps not entirely successfully, but usefully enough to give us insight about what kind of problem we must deal with. As with many other things of interest to us today, we have to go back to the famous cyberneticist and scientific gadfly Norbert Wiener.

Wiener developed a distinction between two kinds of demons that obstructed agency in the world. One demon is “Augustinian.” It is primarily characterized by chance and disorder but nonetheless is incapable of anticipating our actions and altering its strategy. The Augustinian demon can be “revealed” through any process that resolves ambiguity or confusion. Another demon is “Manichean” in that it is capable of adapting its strategies and tactics to deceive, obstruct, and outwit us. The Manichean demon’s malice makes any revelation fragile, temporary, or even potentially deceptive. Carl von Clausewitz said that it is critical that one must know what kind of war they are is fighting. But, more generally, it is critical to know what kind of demon one is fighting.

Augustinian demons are identified with nature “red in tooth and claw.” Manichean demons are purposeful conflict actors. It may seem – at least crudely – that the degree of intentionality is the primary demarcation between the two demons. Yet it is also important to understand that for Wiener intentionality was merely a certain type of relationship between inputs and outputs. How much can be ascribed to a particular entity is a continuous, rather than discrete, determination. Adversariality is probably more important than intentionality. For example, when scientists collaborate together to solve a big problem, they are defined by the quality and quantity of their effort and not necessarily by their mistakes. On the battlefield it does not matter if a particular soldier is skilled or has expensive gear – if he makes even a minor mistake he could be instantly killed. As Wiener observed, the fight against the Augustinian demon prioritizes our best qualities and the fight against the Manichean demon prioritizes our mistakes and weaknesses.

Where does the novel coronavirus fit in Wiener’s demonology? The answer is predictably mixed. The coronavirus cannot inflict as much pain and suffering on us as it could if it were maximally adversarial. In war it is a common tactic to use indirect methods – strategic bombing campaigns against enemy industrial sites, unrestricted submarine warfare against enemy merchant shipping, or commandos raiding enemy rear-area supply depots – to degrade the opponent’s structural capacity to make war. The virus is not capable of maneuvering itself such that it prioritizes the targeting of personnel operating critical medical manufacturing sites, research and development labs, supply networks, and other important material and human resources in order to remove our ability to resist. All of these capabilities may be degraded for a variety of other reasons, but not necessarily out of adversarial design. Disease organisms are directly selected in evolutionary processes to compete against other disease organisms rather than humans.

So yes, the virus is very Augustinian. Especially when one also considers that theoretical and applied scientific and technical advances could conceivably generate lasting advantages that the virus would be hard-pressed to counter. As it is the “novel” coronavirus it has only limited ability to change its nature once that underlying nature is revealed through rational investigation. It still enjoys the element of surprise due to uncertainty about its characteristics and behaviors. That element of surprise can be taken from it, just as other infamous pathogens lost their protective armor of mystery. At the same time, there is also something uncomfortably Manichean about how exploitable even small human mistakes are. This is implicit in the “systemic” risk that the virus poses. A car crash can be lethal for an individual but smaller for society. Individuals may survive coronavirus but at the cost of infecting many others directly and/or indirectly.

Every small relaxation of our guard potentially risks disaster. “Potentially” is the keyword, at least in this climate of relative uncertainty. Given the supernatural language Wiener introduces perhaps it is not particularly surprising that much of the tools we rely on the most to make decisions about the virus are technological prosthetics that promise us the ability to “see” beyond our immediate sensory world. The coupled dynamics of human populations and the virus are not perceptible to us without these prosthetics. They are in part formalizations of informal mental processes we perform intuitively – we all make conjectures about how a complicated system interaction will play out, but not all of these conjectures are explicit or testable. Thus, we rely heavily on scientific modelling to think about the course of the coronavirus.

There are complicated ethical issues with relying on modeling but the uncertainty of the moment and the need for some way to test potential courses of action makes them invaluable. Models are useful because of the inability to directly experiment at scale on policy outcomes (and the dubious ethics of doing so even if it were possible). However, there is no free lunch. The quality of the input data matters a great deal and the more significant the problem the less likely one is to find data that lacks significant complications. And the more complicated the model, the more complicated questions of verification, validation, and testing will be. We should also more specifically note that we are dealing with simulation per se rather than “modeling” as a generic descriptor.

All models are to some extent useful fictions. But not all fictions are created equal. Many coronavirus models are computer simulations, such as the March 16 Imperial College London report that influenced the British government’s changed pandemic policies. What is distinct about computer simulation? Plenty. In broad outline (of a far more complicated topic), a computer simulation involves choosing a model of a target system, implementing it on a computer, studying the output of the implementation, deriving inferences from the output, and then using the inferences to conclude something about the target.

In designing and implementing a computer simulation, the simulator draws from many sources outside of relevant theory by necessity. Intuition and approximation is often necessary to make complicated implementational choices regarding the simulation’s form, function, and parameters. This makes simulations “autonomous” because a simulation often is not perfectly tied to the theory it notionally is drawn from or directly comparable to real world data and observations. Many of these implementational choices make simulations subject to suspicion. A good deal of this criticism misses the mark, but even in the best case we must remember that the simulations exist independently of the “real” world.

The concrete fictions of computer simulations ultimately cannot displace the sense of shock, panic, and depleted agency that Rao eloquently describes. After all, if the problem is psychological and culture shock of the unfamiliar it is unlikely that creating virtual societies and running experiments on them will do much to make the problem more tangible and palatable even if it may aid scientists and policymakers in taming it. From the beginning, the virus’ mixture of Augustinian and Manichean characteristics defies human mental models. Despite the fact that human societies being held hostage by pathogens is a story as old as human civilization itself. After all, the virus is brutally indifferent to the manner in which human societies fib their way out of problems by manipulating social perceptions of them.

Let us briefly return to what was said earlier about the way we regard ourselves as unique and special. There is obviously something to that beyond mere vanity. Human minds can imagine, deliberate, and collaborate. However, by the same token they also frequently confabulate. The former qualities – our abilities to think and act – are what can produce the scientific and public policy efforts that will eventually take us out of our deadly embrace with the novel coronavirus. This is the way in which the virus is an Augustinian demon. But the latter qualities – our limitless faculty for self-deception – are part of what have produced massive death tolls and they showcase the manner in which Manichean demons ultimately define us by our frailties and mistakes. This distinction must not be forgotten, even if the virus does not cleanly fit into a particular demonic category.